ASReview LAB
Systematically screening large volumes of textual data is time-consuming and cognitively demanding. Active learning-based screening prioritization has emerged as a well-established approach to support researchers in identifying relevant records more efficiently, while maintaining transparency and control over the decision-making process.
ASReview LAB is an open-source software environment designed to support AI-aided systematic reviewing by interactively prioritizing records for screening, with the aim of finding all relevant literature while minimizing the number of records that need to be read by humans.
Progress
The ASReview project was first introduced in Nature Machine Intelligence as an active learning framework for systematic reviews. Since then, ASReview LAB has developed into a mature software package with a user interface that supports both regular reviewing projects and methodological research into AI-aided evidence synthesis.
In May 2025, ASReview LAB 2.0 was released, introducing a redesigned interface, improved usability, and a clearer separation between different user workflows. The accompanying paper published in Patterns in July 2025 describes how ASReview LAB 2.0 introduces an advancement in AI-assisted systematic reviewing by enabling collaborative screening with multiple experts (“a crowd of oracles”) using a shared AI model.
A major architectural update in ASReview LAB 2.0 is the separation between lightweight and heavier machine learning models. A lightweight option is provided through the ELAS u4 model, which focuses on fast, efficient classifiers suitable for the majority of research projects. For more advanced use cases, ASReview LAB integrates with the Dory package, which contains heavier machine learning models (ELAS h3) and language models (ELAS l2). You can find the accompanying paper here. Dory is designed for scenarios where richer text representations and more computationally intensive models are required, while keeping the core ASReview LAB environment responsive and usable. This separation allows users to continue screening using a lighter model, while a heavier model trains on previously labeled data in the background.
ASReview LAB continues to be updated and improved with new releases, based on contribution from its users base on the GitHub page. The full history including detailed release notes can be found there, too. The GitHub environment also contains a Discussions page where a lively community of users share their experiences and workflows. Lastly, ASReview has its own website, where all tutorials, extensive documentation and updates on future events can be found.
Funding
This project is funded as part of prof. dr. Rens van de Schoot’s VICI project, titled Transparent and Reproducible AI-aided systematic reviewing for the Social Sciences (TRASS), funded by the Dutch Research Council (VI.C.231.102).
People involved
- Timo van der Kuil – PhD Candidate
- Jelle Teijema – PhD Candidate
- Peter Lombaers – Engineer
- Yongchao Terry Ma – Engineer
- Berke Yazan – Frontend Developer
- Angie Dong – UX Developer
- Jonathan de Bruin – Lead Engineer
- Rens van de Schoot – Project Lead